Pull Request Explorer

Summary

Add visible divider to queued messages

Improves UI clarity by adding a subtle divider between queued messages, enhancing user experience in dark themes.

Health Assessment

• • Quick AI‑assisted style fix with minimal code changes and no rework, merged within 12 minutes.

Summary

Fix Slack Connect Error and Slash Commands

Resolves issues with Slack integration, improving user experience.

Health Assessment

• • Quick merge with minimal review time, indicating a straightforward fix.

Summary

Add self-review proposal actions to briefs

Enables agents to automatically propose and execute actions on briefs during self-review, improving automation and reducing manual intervention.

Health Assessment

• • PR merged in 1.6 hours with only two review comments, indicating high confidence; the change is large but heavily tested and reviewed by an AI bot, reducing human review friction.

Summary

Polish confirmation bar layout

Improves user experience by tightening spacing and aligning UI elements in the intervention bar, making actions clearer and reducing visual clutter.

Health Assessment

• • Merged within 20 minutes with a single review comment, indicating minimal friction and a small UI polish change.

Summary

Refine homepage banner copy for channels and skills

Ensures consistent banner copy across English and Chinese locales, improving user experience and reducing confusion.

Health Assessment

• • Quick review and merge with minimal friction, small scope.

Summary

Fix SKILL.md frontmatter parsing and validation

Improves skill document editing by ensuring accurate frontmatter parsing and validation, preventing errors and enhancing user experience.

Health Assessment

• • Rapid 4‑hour cycle with AI‑assisted review, but the large number of added lines indicates a substantial refactor and feature enhancement.

Summary

Add 4h JWT for hetero-agent operations

Extends JWT expiry to 4 hours for hetero-agent tasks, preventing 401 errors and improving reliability for long-running AI operations.

Health Assessment

• • Merged within 0.7 hours with minimal review, indicating a straightforward change with low complexity.

Summary

Release 20260509 with new features

This release introduces the Agent Task System GA, nightly self‑review, inline documentation, and expanded bot platforms, enhancing automation and user experience.

Health Assessment

• • Rapid 4‑hour cycle with many commits indicates extensive iteration; large scope and numerous changes raise risk despite short review time.

Summary

Remove image avatar from lab input markdown

This cosmetic change removes the avatar image from the lab input markdown rendering item in Settings > Advanced, ensuring visual consistency across all lab items.

Health Assessment

• • Quick cosmetic change merged within minutes with minimal review.

Summary

Fix task agent manager behavior

Improves task agent manager stability and UI, ensuring correct agent scoping and display across task views.

Health Assessment

• • Rapid iteration with minimal review; large code changes but no blockers.

Summary

Migrate Notion to LobeHub Market

Migrates the Notion integration into the LobeHub marketplace, streamlining skill management and enhancing user experience by consolidating provider data and improving task templates.

Health Assessment

• • Fast review and merge with minimal iterations, indicating high confidence in the change.

Summary

Recover malformed tool_call names instead of silent finish

Fixes LLM tool call handling by recovering names without prefix, adding diagnostics, and improving reliability for the agent runtime.

Health Assessment

• • Merged in 1.2 hours with minimal review, indicating a straightforward bug fix and low complexity.

Summary

Add Cloud Claude Code V3 repo picker and sandbox

Enables users to select GitHub repositories for Claude Code execution, injects a GitHub token into the sandbox, and adds sandbox context handling, improving integration and security for cloud-based agent workflows.

Health Assessment

• • The PR was merged quickly after a single review, but required multiple small commits to address feedback, indicating iterative refinement. The large code addition and many files changed raise the risk profile despite the short cycle time.

Summary

Enforce git-derived PR refs and metrics

Hardens the version-release skill to derive release‑note data from git, ensuring accurate PR references and metrics and reducing manual errors.

Health Assessment

• • Fast cycle time (1.1h) and immediate review comments indicate low complexity and high confidence in the documentation changes.

Summary

Fix signin captcha flow removal

Removes unused CAPTCHA widget from sign‑in, preventing unnecessary errors and improving user experience.

Health Assessment

• • Quick fix with AI review, minimal rework, fast merge.

Summary

Fix Docker build by using static package.json

Ensures Docker builds succeed by preventing corepack validation errors caused by pnpm init.

Health Assessment

• • Quick fix with minimal changes, no rework.

Summary

Fix temporary email auth error locale

Adds a localized error message for temporary email authentication, improving user experience by providing clear, localized warnings instead of raw backend messages.

Health Assessment

• • Fast cycle time and minimal review comments indicate a straightforward change with quick approval.

Summary

Add bot callback service

Fixes bot callback handling, improving reliability of bot interactions.

Health Assessment

• • PR merged quickly with minimal review, indicating a straightforward bug fix.

Summary

Sanitize sensitive comments in production JS bundle

Removes hardcoded secrets and example URLs from bundled JavaScript to mitigate information disclosure risks. Ensures compliance with security standards and protects user data.

Health Assessment

• • Quick fix with minimal changes and immediate review, indicating low complexity and high confidence in security patch.

Summary

Add stop run action and proxy matcher

Users can now cancel running tasks directly from the activity card, and task routes no longer 404 in production.

Health Assessment

• • Fast review and merge with minimal rework; small code changes; low risk.

Summary

Update auth captcha retry copy

Improves user experience by clarifying captcha retry instructions across multiple locales.

Health Assessment

• • Fast review and merge with minimal changes indicates low complexity and risk.

Summary

Fix multiple account link

Corrects account linking logic to prevent duplicate or mismatched account associations, improving user experience and data integrity.

Health Assessment

• • Rapid resolution with minimal review, indicating straightforward config updates.

Summary

Fix onboarding footer visibility

The onboarding footer is now only visible when specific conditions are met, reducing inappropriate displays.

Health Assessment

• • The PR was reviewed and merged quickly, indicating a straightforward fix.

Summary

Refine Agent Signal Receipt Cards

Enhances user experience by making agent signal receipts more accessible and interactive, reducing confusion and improving engagement.

Health Assessment

• • Rapid review and merge with minimal changes indicates high confidence and low risk.

Summary

Set OSS default model to DeepSeek V4 Pro

Updates the open‑source default LLM to DeepSeek V4 Pro, tightening regression tests and ensuring consistent model selection.

Health Assessment

• • Fast review cycle with AI review bot, minimal rework, quick merge within 2 hours.

Summary

Fix hidden runtime-only model aliases

Ensures runtime-only model aliases are hidden from user-facing lists, improving clarity and preventing accidental exposure.

Health Assessment

• • Quick turnaround with minimal review comments indicates a straightforward bug fix.

Summary

Remove insecure proxy endpoint and dead plugin code

The PR eliminates an unauthenticated open HTTP proxy that could be abused as a relay for external resources. It also deletes dead URL‑manifest plugin code, reducing attack surface and code complexity.

Health Assessment

• • Fast turnaround with minimal review, indicating a straightforward security fix; large code removal but no major functional changes.

Summary

Health Assessment

Summary

Fix hetero persistence state across Vercel replicas

Ensures consistent state for heterogeneous agent across Vercel serverless replicas, preventing message fragmentation and orphaned tool calls, improving reliability for multi-step tasks.

Health Assessment

• • Quick resolution with minimal rework; AI-assisted code generation streamlined fix.

Summary

Fix agent skill folder display and delete confirm

Improves user experience by correctly rendering managed skill bundles as folders, showing SKILL.md content, and aligning delete confirmations with app style, reducing confusion and errors.

Health Assessment

• • Rapid iteration with AI review bots and multiple commits after the first review led to a large code churn, indicating significant refactoring of agent document handling.

Summary

Refactor onboarding to shared prefix steps

Extract language and privacy as shared prefix steps in onboarding flow, improving user experience and reducing complexity.

Health Assessment

• • The PR has a large scope with many files changed, but the review process was relatively smooth with no major issues raised.

Summary

Agent Task System General Availability

Enables the Agent Task System for all users, removes the feature flag, fixes task creation flow, and adds UI polish.

Health Assessment

• • Fast cycle time and minimal review iterations indicate efficient collaboration; AI assistance accelerated development.

Summary

Fix generation moderation context passing

Ensures moderation context is correctly passed during image and video generation, improving error handling and user feedback.

Health Assessment

• • Quick turnaround with minimal review friction; AI review bots provided suggestions but no major rework was required.

Summary

Fix DeepSeek reasoning_content in OpenAI compatibility

Ensures DeepSeek models correctly handle reasoning content, preventing API errors and enabling tool usage in OpenAI-compatible runtimes.

Health Assessment

• • Rapid resolution with minimal rework; AI-generated code was reviewed quickly and merged within 30 minutes.

Summary

Add nightly self-review feature for agent signals

Enables automated nightly self-review of agent signals, improving data quality and reducing manual oversight.

Health Assessment

• • Rapid merge with minimal review indicates high confidence, but the large code change size raises potential risk for downstream impact.

Summary

Improve signal receipt retry with exponential backoff

Enhances reliability of signal receipt fetching, reducing failures and improving user experience.

Health Assessment

• • Quick review and merge within 1.6 hours indicates smooth process.

Summary

Fix missing document tool outcome rendering

Ensures agent documents correctly emit tool outcome events, improving agent decision accuracy and client runtime hydration.

Health Assessment

• • Rapid review and merge with minimal back‑and‑forth, but the change touches many files and lines, indicating a high‑impact bug fix that required extensive testing. The presence of AI‑powered review tools suggests automated quality checks were leveraged.

Summary

Style: Fill input on follow-up chip click

Allows users to edit suggested follow-up messages before sending, improving usability and reducing accidental sends.

Health Assessment

• • Fast cycle time and minimal rework indicate a smooth review process; automated AI review bots provided quick feedback without blocking the merge.

Summary

Fix prompt cache key for OpenAI chat requests

Adds a prompt cache key to improve caching and error handling for OpenAI chat requests, enhancing performance and reliability.

Health Assessment

• • The PR underwent multiple AI‑assisted reviews and required several iterations, resulting in a prolonged cycle time. Despite the large number of lines added, the scope remained moderate, and the use of AI review tools helped identify issues early.

Summary

Batch fork API for parallel marketplace install

Parallelizes agent fork calls during marketplace onboarding, cutting round‑trips and speeding up user setup.

Health Assessment

• • Rapid merge with minimal review indicates high confidence and low complexity.

Summary

Refactor: implement cloud hetero agent exec pipeline

Adds a cloud‑based heterogeneous agent execution pipeline, enabling server‑side orchestration and CLI integration for improved scalability and session management.

Health Assessment

• • Rapid AI‑assisted iteration produced a quick 31‑hour cycle, but the large scope and multiple post‑review commits suggest careful testing is required.

Summary

Fix MentionMenu scroll clipping in chat input

This change resolves visual clipping of the MentionMenu dropdown in the chat input, improving user experience by ensuring smooth scrolling and eliminating border artifacts.

Health Assessment

• • Quick, single-file fix with no tests added; minimal review friction and low risk.

Summary

Polish onboarding UI and add tool result renders

Enhances onboarding experience with refined UI components and richer tool result displays, improving usability and clarity for users.

Health Assessment

• • Rapid iteration with minimal review time indicates high confidence in changes; UI polish delivered quickly with a single review cycle.

Summary

Add agent Hono routes

Adds new API endpoints for agent integration, enabling messenger install, OAuth callback, and webhook handling to expand platform capabilities.

Health Assessment

• • Fast turnaround with single commit and minimal review, indicating low complexity and high confidence in the changes.

Summary

Fix brief/task-template fetch overhead on home

Reduces unnecessary network calls and improves home page load times, enhancing user experience.

Health Assessment

• • Fast review cycle and minimal rework indicate a well‑structured, low‑risk change that improves performance without altering core functionality.

Summary

Add comment tools and reparent support

Enables users to comment on tasks and move tasks within a hierarchy, improving collaboration and task management.

Health Assessment

• • Fast turnaround with minimal review iterations indicates high confidence in code quality.

Summary

Fix user field preservation on session refetch

Ensures user profile data persists across tab focus refetches, improving UI stability and reducing unnecessary network calls.

Health Assessment

• • Quick fix with minimal changes, reviewed by AI tools, merged within an hour.

Summary

Fix telegram API timeout handling

Corrects telegram bot callback errors and adds timeout handling, improving reliability for users interacting with the chat bot.

Health Assessment

• • Quick turnaround with minimal review, indicating a straightforward bug fix that resolved a production issue.

Summary

Fix skill intent handling and signal routing

Improves agent signal processing by correctly handling skill intents when hintIsSkill is enabled, ensuring reliable skill triggers and preventing misrouted signals.

Health Assessment

• • Rapid iteration with minimal review comments; large code churn indicates significant refactoring of agent signal logic.

Summary

Fix support provider sdk type routing

Adds a generic router-runtime option field to allow provider channels to select an SDK wrapper, improving routing for DeepSeek and Moonshot providers.

Health Assessment

• • The PR was merged quickly with minimal review friction, indicating a well-structured and necessary change.